The paper delivers the first comprehensive systematization of adversarial robustness in QML with new empirical tests showing an accuracy-robustness trade-off, amplitude encoding's vulnerability, and QML's greater susceptibility to evasion attacks than classical models.
Towards defending crosstalk-mediated attacks in multi-tenant quantum computing
1 Pith paper cite this work. Polarity classification is still indexing.
abstract
With the increasing demand for quantum hardware, shared and multi-tenant environments have been proposed to optimize resource utilization. However, the multi-tenancy paradigm in quantum computing inherently introduces security threats. This paper examines crosstalk-mediated attacks targeting three-qubit Grover's search algorithm and explores two fundamental mitigation strategies: gate-based dynamical decoupling and the use of a buffer qubit. We evaluate the effectiveness of each method individually and in combination, finding that while both strategies offer some level of attack mitigation, their combined application yields the most significant performance improvement. Beyond security vulnerabilities, our work also has implications for unintentional circuit interference that can occur when multiple quantum circuits are executed in close proximity.
fields
cs.CR 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
SoK: Critical Evaluation of Quantum Machine Learning for Adversarial Robustness
The paper delivers the first comprehensive systematization of adversarial robustness in QML with new empirical tests showing an accuracy-robustness trade-off, amplitude encoding's vulnerability, and QML's greater susceptibility to evasion attacks than classical models.